A third of all food produced worldwide is lost or wasted every year according to the United Nations Food and Agriculture Organization. Losses for perishable goods represent an economic drain for the entire world, but for individual companies it can be even worse: spoilage of perishable goods like food and medicine can put a company and its corporate officers at legal risk.
At SAP TechEd we will demonstrate how intelligent containers powered by SAP HANA® can tackle challenges like this outside of the data center during product storage and transportation.
Intelligent containers can help secure goods and provide real-time monitoring of freshness while also using predictive analytics to project product quality. Our demo uses SAP HANA, express edition to couple the power of the SAP HANA platform with mobility to enable data aggregation and local insights in network-edge deployments. This demo shows how SAP HANA data platform alone can support all different types of data and predictive analysis. The analytic dashboard is a native SAP HANA web application, which simplifies the application development significantly.
As you can see in the following diagram, each product is held on a transparent base with a set of sensors to monitor the product’s weight, temperature, and humidity. With the real-time product information, we can actively track and predict the product quality based on the historic data, to help businesses make better decisions.
This demo showcases the power of SAP HANA to handle several different types of data simultaneously:
- Streaming analytics: The demo streams real-time data into a time-series forecasting model from nearly 100 sensors (generating 140 data points a minute)! The sensors are in and on the container and on the 20 product holders inside the intelligent container. The demo also uses SAP HANA streaming analytics to filter out anomalous data and to generate alerts based off predetermined baseline values for sensor data. Beyond global anomalies, the demo also uses real-time streaming analytics in SAP HANA to merge sensor data from multiple input streams to provide real-time status monitoring for specific products stored in the container.
- Predictive analysis: Data streamed from the numerous sensors (RFID for location, FSR- force sensing resistors for weight and BLE temperature sensors) also help the demo predict product health. Real-time sensor data is fed into a time-series forecasting model to infer when a product will reach the optimum temperature for consumption and when a product will expire. The demo uses a time-series Automated Predictive Library (APL) model that is that is accessed via a SQL call every five minutes. The model takes into account the current temperature of a given item, the item’s specific location within the container, and its weight to predict when the product will reach its optimum temperature for consumption (12 °C), regardless of where it is placed or how much it weighs.
- Geospatial data: The demo uses geospatial data to track products stored in the intelligent container for deviations from three planned transportation routes leading to Las Vegas, Orlando, and San Francisco. Deviations from these planned routes are important because they signal changes in delivery time that could endanger product health or physical security. Geospatial point-data in the demo is fed to the Bing Map API where an initial route is plotted; the actual route location is plotted to the map in real-time to replay the trip progress with locational route data, with each route containing between 350 and 1300 individual data points.
- Open-interface integration with TensorFlow: The demo couples its real-time sensor data with image recognition in order to track inventory. The container tracks the location of every item within it from the moment the item is put in the container using sensor data; an image-recognition algorithm built in TensorFlow tracks when a particular item is placed into the container and when it is taken out. The image-classification model classifies items with a percentage of certainty. The model was trained on 300-500 images of each type of soda can used in the demonstration (Coca-Cola, Sprite, and Fanta). When the classifier identifies one of the types on soda, the streaming analytics create an alert and the system associates the image with the next item to be placed within the fridge so that the system can recognize when the item is taken out in the future.
SAP engaged Prowess Consulting to create this demo. Founded in 2003, Prowess specializes in working with Fortune 100 tech leaders to define, manage, and market solutions and services. As part of their unique process, they fuse experience from years of marketing strategy and creative storytelling with the features and functions of their customers’ advanced technologies to ignite explosive, rare power in every engagement.
While this particular demonstration highlights how real-time and predictive analytics can help secure global cold-chains for perishable goods, the power of these analytics moved out of the data center and onto edge devices can benefit any industry that has to move, monitor, or secure high-value goods. For example:
- Hospitals: prevent drug diversion & spoilage
- Food Vendors: mitigate risk of food poisoning
- Suppliers: track shipments and live inventory health
- Controlled Beverages (i.e. booze): gate and analyze access
- Primary Care: ensure vaccinations at correct temperatures
So come by the genius bar at the SAP experience area and see the SAP HANA intelligent container – and the Internet of Things – in action at SAP TechEd.
 Source: Food and Agriculture Organization of the United Nations. 2011. Gustavsson, J., Cederberg, C., Sonesson, U., Otterdijk, R.v., & Meybeck, A. “Global Food Losses and Food Waste: Extent, Causes and Prevention.” Food and Agriculture Organization of the United Nations and Swedish Institute for Food and Biotechnology (SIK). www.fao.org/docrep/014/mb060e/mb060e00.pdf. Reproduced with permission.